[clang] [AMDGPU] Modifies mfma builtin def to take _Float16 for HIP/C++ (PR #175197)

[Shilei Tian via cfe-commits]

================
@@ -0,0 +1,339 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 6
+// REQUIRES: amdgpu-registered-target
+// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -target-cpu gfx908 -DMFMA_GFX908_TESTS -emit-llvm -fcuda-is-device -o - %s | FileCheck %s --check-prefix=CHECK-GFX908
+// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -target-cpu gfx942 -DMFMA_GFX942_TESTS -emit-llvm -fcuda-is-device -o - %s | FileCheck %s --check-prefix=CHECK-GFX942
----------------
shiltian wrote:

The convention is to add gfx suffix to the file, like `clang/test/CodeGenHIP/builtins-amdgcn-mfma-gfx908.hip`, `clang/test/CodeGenHIP/builtins-amdgcn-mfma-gfx950.hip`.

You are only touching builtins with two different target features, so I expect you only need to test two targets, aka two files.

https://github.com/llvm/llvm-project/pull/175197